2015-02-10

VMware Integrated OpenStack - Cost Analysis

VMware announced last week the launch of VIO and there are a number of things that I think people are missing and should be pointed out.

The information I have taken is from the Datasheet and publicly available information.

Networking

A great part of the the functionality and flexibility that people use is the option for flexible networking, i.e. creating private networks, routers for example.

NSX for Neutron

That is great – all the functionality is there – with NSX. But how many people are actually using NSX today? How many people have deployed NSX? In a previous article I went through the reasons why this will not be an easy path. So how does that work with what I currently have in my datacenter?

I think it is safe to assume that this will be deployed in an environment with a Distributed virtual switch – we are talking about environments that are using Enterprise Plus after all (and VMware is giving this away for free to everyone with the Ent+ licenses).

So how does OpenStack work with a DvSwitch today?

Well I am sorry to surprise you – but it does not. It only works with nova-network (which is supposed to be deprecated – so caveat emptor). VMware themselves have said that most customers that are using OpenStack and VMware are using NSX (and that they don’t really have much experience with nova-network).

nova-network???

** Edited February 11th, 2015 **

According to a twitter conversations with @hui_kenneth and @danwendlandt last night – VIO GA will support the dvSwitch, only that information is currently not public and is only available to the Beta users. The functionality still will not bet the same as that of NSX.

So it boils down to this. The only way to really use OpenStack with vSphere today – in any kind of semi-normal way, is to do it with NSX. Any demo you have seen – Hands-on Labs, presentations all use NSX. And it always something that already exists in the environment you are working with, that is the assumption.

So VMware is giving this away for free (unless you are interested in support – which will cost you another $200 per socket) – but this essentially is giving you a hobbled product – which does not have functionality that you get out of the OpenStack box – because you are using vSphere networking.

So what features will you not be able to use – without NSX?

  • No GRE – standard VLANs only
  • No LBaaS
  • No VPNaaS
  • No FWaaS
  • No security groups

I will say that the number of people that are actually using FWaaS and VPNaaS are not the majority of OpenStack users – but on the other LBaaS – is more or less an essential part of any automated cloud. And even more so – security groups are definitely an essential part of any cloud.

But of course we would like to use all the bells and whistles – actually I would really only like to use Neutron with vSphere – so my options are only going to be to use NSX (until they manage to get this working as will with a dvSwitch).

So what is this going to cost me?

This is going to hurt (and by no means am I licensing expert – and yes I know that no-one really pays list price – but here goes).

You have two options to buy NSX – per vm or per socket. Now we all know that the per-vm model – usually does not run in the customers favor – and the last time a per-vm model was proposed – the was a huge disturbance in the force. So I am going to assume that you will want a per CPU based license.

1 CPU license of NSX for vSphere (NS-VS-C) – $5,996
1 CPU SNS Basic support (NX-VS-G-SSS-C) for 1 CPU – $1,259
1 Year SNS Production support for VMware Integrated OpenStack for 1 CPU – $200
1 CPU license of VMware Integrated OpenStack (assuming you have Enterprise Plus) – $0

Total cost for 1 CPU of VMware Integrated OpenStack – $5,996
Annual support costs for 1 CPU of VMware Integrated OpenStack (and NSX) – $1,459

So let me lay this out in simple terms with an example.

** Post Updated February 11th **

It was brought to my attention that there is a minimum purchase of 50 CPU licenses for OpenStack support as part of the FAQ notes.

Minimum of 50

I did not change my assumption that you would only be using 4 hosts but the purchase of additional OpenStack CPU support s required.

I have therefore amended the numbers below.

If you are interested in really using (i.e. with neutron and NSX) OpenStack on a 4 host (8 socket) cluster this will cost you:

  • Initial cost
    • NSX licensing – 8x$5,996 = $47,968
    • SnS first year – 8x$(1,259 + 200) = $11,672
    • SnS first year – 8x1,259 + 50x200 = $20,072
    • Total – $59,640
    • Total – $68,040
  • Annual costs
    • SnS per year – $11,672
    • SnS per year – $20,072

That is above and beyond the regular licensing fees that you pay for Enterprise Plus licenses (which I have not factored in here – because I am assuming that you already have them. But if you do not, then that is even more of a hit to your CAPEX.

Again I would like to stress – that this is MSRP – and not including any bundles.

My Take

VMware would like to see the whole world run on their platform (obviously), and they have started to make a move to minimize the impact that I think they are starting to feel – due to people moving over OpenStack. This offering is a foot in the door to minimize the business they could lose from people moving off of their platform (seriously speaking though vCloud is a competing product – and I do not know how much longer they can continue to sell competing products). There are a number of benefits of running on top of vSphere of course, the underlying platform – and the hooks and insight into vRO is another one of course.

That is one side of the story. The other side is NSX adoption – I do not think that VMware is seeing the market share that they were hoping to gain with NSX – network virtualization is still not a mainstream concept. Companies are starting to dabble and try – but no – we are not there yet.

openstack

The ironic thing is that even with VIO integrated with NSX when it is released – it still will not support native LBaaS, VPNaaS and FWaaS out of the box (you probably will be able to integrate with 3rd party vendors) – that will probably come in a future release.

So even with their flagship product – it still will not have all the functionality that OpenStack Operators/Users are accustomed to have in their environments today.

There are benefits of having “one neck to throttle” so to speak – but that comes with a price tag – and hefty one. It certainly is not a free product as it is being made out to be – or at a minimal cost (VMware support).

The devil is always in the details.

What do you say? Is it financially viable? Would you use VIO? Why? Or would you rather rough it and go with another vendor?

I would be happy to hear your thoughts, and comments. Please feel free to leave them in the comments below.

2015-01-21

The OpenStack Elections - Another Look

The Board of directors and the bylaws were approved. Summary posts can be found here (2015 Individual Director Election results) and here (Bylaws amendments approved).

Individual Directors

  • Tim Bell
  • Russell Bryant
  • Alex Freedland
  • Rob Hirschfeld
  • Vishvananda Ishaya
  • Kavit Munshi
  • Egle Sigler
  • Monty Taylor

The voting numbers can be found here.

Congratulations to all the new and re-elected board members. Well deserved!

I have a few things I would like to add about the data that was presented – and my thoughts.

1. Tim Bell 

Tim received the highest number of votes. CERN is a huge OpenStack user and was showcased at the summit in Paris. He is one of only 3 the board members that are not officially affiliated with a vendor directly involved in OpenStack. I see this as a big vote of confidence by the members of the foundation – that are interested in seeing more representation from non-affiliated members. Something I personally would like to see as well.

2. Participation Numbers

2015-01-21_04-22-06

Only just over 16% of the members voted in the election. That to me seems to be very low. and I would like to address the OpenStack community to ask why this is so? This is actually something the Board and Foundation should also be actively looking into as well (perhaps they already are).

  • Is it because people are not interested in participating?
  • Is it because the importance of the process was not made clear enough?
  • People did not find the candidates suitable?

More people actually voted for the amendment changes that in the election itself – which I find quite strange. They were already on the page and did not bother to vote for any of the candidates.

3. Operators were not chosen

Neither Jesse Proudman and Randy Bias were voted in – both seemed to me that they were very vocal in their campaign as to why they wanted to get on the board – and that was to bring a change into the way OpenStack is currently “run”. I do personally think this is pity – because I would have liked to see more representation from the Operator standpoint, something which I still feel is still lacking in the OpenStack community.

4. Participation in Amendment changes

2015-01-21_04-30-49

The quorum was achieved – but only by 315 votes. My previous post The OpenStack Foundation – 2015 Individual Director Election – spoke about how I see this change as a problematic one. I personally do not think that the quorum would have been reached – if it was not for the “aggressive” marketing campaign that the Foundation embarked upon in order to reach this quorum. Without the countless number of posts from Board members, Foundation members (myself included) and anyone that cared about this election - on their blogs, social media and everywhere that was possible. Jonathan Bryce even promised to remove his beard to achieve this goal..

And he did!

A huge effort indeed, from the foundation and the community at large – but my question is..

How long will it be until it is needed again?

Yes the quorum needed is now officially only 10% (instead of 25%) but I do foresee the day that even that will not be reached (and that will not be too far in the future). That is why I think the change should have been to remove the quorum all together.

5. People were not happy with the change in the quorum

On both the first two amendments, the approval rate was 93-95% – almost everyone agreed. Changing the quorum – only had 80% of the people that approved. Of course that is still a majority and perfectly valid and acceptable as a decision, but still it is interesting to see that more people were not happy with the change.

It would be interesting to know if it was the lowering of the percentage to 10% or was it that the proposed change should have been to remove the quorum altogether. I personally voted against this change because I think that the quorum should be removed completely and not lowered to 10%.

I am very pleased that the changes were made – because it allows OpenStack to continue to grow, but I do think that planning should start now – for when the changes made in these amendments will not suffice and need to be changed again.

If you are willing to share - I would be interested in hearing your thoughts about the points above. Please feel free to leave them in the comments below.